Document not found! Please try again

Fragment Based Drug Discovery: Practical Implementation Based on

Dec 13, 2011 - Here, we demonstrate a more general use of 19F NMR-based fragment screening in several areas: as a key tool for rapid and sensitive det...
0 downloads 6 Views 3MB Size
Article pubs.acs.org/jmc

Fragment Based Drug Discovery: Practical Implementation Based on 19 F NMR Spectroscopy John B. Jordan,*,† Leszek Poppe,*,† Xiaoyang Xia,† Alan C. Cheng,† Yax Sun,† Klaus Michelsen,† Heather Eastwood,† Paul D. Schnier,† Thomas Nixey,‡ and Wenge Zhong‡ †

Department of Molecular Structure, and ‡Chemistry Research and Discovery, Amgen, Inc., One Amgen Center Drive, Thousand Oaks, California 91320, United States S Supporting Information *

ABSTRACT: Fragment based drug discovery (FBDD) is a widely used tool for discovering novel therapeutics. NMR is a powerful means for implementing FBDD, and several approaches have been proposed utilizing 1H−15N heteronuclear single quantum coherence (HSQC) as well as onedimensional 1H and 19F NMR to screen compound mixtures against a target of interest. While protonbased NMR methods of fragment screening (FBS) have been well documented and are widely used, the use of 19F detection in FBS has been only recently introduced (Vulpetti et al. J. Am. Chem. Soc. 2009, 131 (36), 12949−12959) with the aim of targeting “fluorophilic” sites in proteins. Here, we demonstrate a more general use of 19F NMR-based fragment screening in several areas: as a key tool for rapid and sensitive detection of fragment hits, as a method for the rapid development of structure− activity relationship (SAR) on the hit-to-lead path using in-house libraries and/or commercially available compounds, and as a quick and efficient means of assessing target druggability.



INTRODUCTION Fragment based drug discovery (FBDD) has become a major tool for discovering novel therapeutics.1 In contrast to traditional screening approaches, fragment based screening (FBS) relies on the identification of small, weakly potent compounds with good binding efficiencies as a chemical starting point. There are several commonly cited benefits to FBDD: (1) sampling of higher chemical diversity, (2) higher hit rates, and (3) higher efficiency of binding.2−5 Because of the small size and low chemical complexity, fragment libraries can sample a higher proportion of chemical space.2 In addition, the decrease in complexity may result in less constrained, more optimal interactions with the target, thereby increasing ligand efficiency (LE).6 Ligand efficiency is an important metric to assess the quality of fragments hit compounds and is characterized by the binding energy per heavy atom (HA) (eq 1).

LE = −

ΔG #HA

in various phases of clinical development that originated from fragment-based screening campaigns.7 In addition to the widespread implementation of FBDD, there has been significant development in the analytical technologies used to perform FBS. The vast majority of fragment screening is performed using nuclear magnetic resonance (NMR) spectroscopy with proton detection and surface plasmon resonance (SPR) spectroscopy. In the present work, we will focus on NMR and, in particular, will discuss fragment screening methods based on 19F NMR that offer several analytical advantages including speed, sensitivity, and selectivity. Exploiting the fluorine nucleus for NMR screening has been extensively reviewed.8 While the use of the 19F nucleus has previously been proposed for NMR-based screening, these methods have been primarily focused on the screening of more drug-like molecules9 and/or the use of 19F-labeled molecules in reporter screening.8 In the case of a 19F library composed of drug-like compounds,9 compound complexity and poor solubility rendered cocrystallization of the hits with the target highly problematic and posed a high barrier for chemistry follow-up. Recently, Vulpetti et al. proposed the construction of a fluorine fragment library based on the local environment of fluorine (LEF) principle.10 This library, as described, was developed on the observation that the presence of a fluorine nucleus can be beneficial to some properties of small molecule drugs.10−14 Thus, a novel library was constructed to search for the presence of “fluorophilic” protein environments and for identifying binders with fluorine in their pharmacophore.10

(1)

Another benefit of FBDD is that it is typically easier to elaborate fragments into larger, more potent compounds, while maintaining good LE (which typically decreases with increasing MW) compared with traditional high-throughput screening (HTS) hits. While HTS is still used abundantly, FBDD is gaining momentum in the drug discovery process by facilitating the identification of novel low molecular weight (70% similarity was found, we assumed that the reference compound in “fragment space” was represented in the final library (Figures 1 and 2). Through these calculations, it was found that with as few as ∼1200 compounds, approximately 30% of the defined “19F fragment space” was covered. These compounds were acquired with the idea that we could continually acquire more compounds in order to more thoroughly cover the remaining 70% of “fragment space”. Execution of 19F Fragment Screen. The initial fragment stock solutions were prepared at 20 mM stock concentration in nondeuterated DMSO and plated into 96-well plates. The plates containing the fragment library (∼13 plates in all) were pooled by collapsing the plates into one final pool plate, yielding 12−13 compounds per pool. NMR samples of 600 μL in volume and prepared in 96, 1 mL well format were transferred into 5 mm diameter and 10 cm long NMR tubes using a Gilson GX-280 (Gilson, Middleton, WI) and subsequently placed into a refrigerated SampleJet (Bruker Biospin, Billerica, MA) with a temperature setting of 279 K. The reference plate (no protein) and the screening plate were both run under the same conditions for each screen. In principle, if the screening conditions are chosen to be identical for all screens, it should be sufficient to record the reference plate just once. However in practice, hit deconvolution is much easier by having recorded both plates prepared from the same DMSO stock solutions from each pool. This is mainly due to the high sensitivity of 19F chemical shifts toward small mismatches of the DMSO content. Also, possible compound deterioration over time may complicate interpretation of spectra. Using this screening format (in 5 mm NMR tubes), the amount of protein used to perform the fragment screen was approximately 5.0 × 10−7 moles, which for BACE-1 (a ∼47 kDa target) required approximately 20 mg of unlabeled protein. Determination of Kd values. In the limit of [L] ≫ [P] (i.e., small fraction of the bound ligand (pb ≪ 1)), the difference between the observed frequency shift in the presence and in the absence of a protein is described by the following equation:41 vobs − vfree = p b × Δv app (2)

the target can be attributed to nonspecific binding, and these compounds can be excluded from further experiments. In using our 19F fragment library, we have not observed any propensity toward aggregation, as all the compounds were prescreened for a minimum of 1 mM aqueous solubility using 19F NMR. In addition, to test if the binding to a target has 1:1 stoichiometry, Kd values can be calculated from dCSP experiments recorded at different pairs of concentrations. If the Kd values from the two sets of concentrations are consistent, the binding is likely specific to a single site. If the Kd values are inconsistent, then the binding could be nonspecific or the compound could bind to multiple sites with different Kd values. Although the fraction of chemical space covered by fluorine containing fragments may be limited compared to nonfluorinated fragments, our experience with multiple screens using this library suggests that similar hit rates are obtained using both methods. Our method focuses on the use of the fluorine nucleus as a detection tool, and subsequent to a 19F fragment screen, protonated analogues of hit compounds can be mined and pursued in initial hit expansion. This approach can significantly add to the chemical space explored and can provide valuable information and early stage SAR to chemists. In addition, we propose the use of hit rates from screens with the 19 F fragment library as predictors of target druggability. Using targets from a wide range of biological processes, we demonstrate that 19F FBS hit rates correspond very well with commonly accepted computational methods of assessing target druggability and provide an efficient tool for assessing protein topology in the absence of its structure. Taken together, our results suggest that the ability to rapidly perform fragment screens and obtain subsequent SAR using cryogenic NMR detection of the 19F nucleus can provide significant benefits over traditional proton detected techniques.



MATERIALS AND METHODS

Construction of the 19F-Fragment Library. The 19F fragment library was constructed using both proprietary and commercially available compounds. A list of possible compounds was obtained by searching the internal sample bank as well as the ACD database (∼675 000 compounds) using the criteria depicted in Table 1. After application of these filters, an additional diversity selection filter was applied using Daylight Fingerprint (Daylight Chemical Information Systems, Laguna Niguel, CA) as the descriptor. A Tanimoto distance39 of 0.25 was used as the cutoff to ensure the selected compounds were sufficiently dissimilar. After all computational filtering, approximately 1084 internal and 1200 external compounds were identified for possible inclusion in the library. Each of the 2284 compounds was then assessed for 1 mM aqueous solubility using 19F NMR. The purpose of this exercise was 2-fold: First, though screens would be conducted at low concentrations (typically 10−20 μM), we wanted to ensure that soluble compounds made up the library in order to be able to pool a large number of compounds (10−20) and to facilitate X-ray crystallography follow up. Second, we used these experiments to build the database of 19 F chemical shift information which allowed for instant identification of screening hits. Briefly, stock solutions of each compound were prepared at 20 mM in nondeuterated DMSO followed by preparation of aqueous solutions (20 mM sodium phosphate, pH 7.0, 50 mM NaCl) consisting of 250 μM and 1 mM concentrations of each compound. The 19F NMR spectrum of each compound was measured at each concentration, and, through automated processing, the intensities of each spectrum were compared. If the intensity of the 1 mM spectrum was not 4 times (±15%) that of the 250 μM sample, the compound was excluded from the final library. After the solubility assay, approximately 1200 compounds (∼50%) were identified as suitable candidates for inclusion into this library.

where Δνapp is a complex function of the nuclear and kinetic parameters. Importantly, this equation adequately quantifies frequency shift changes for both fast (koff ≫ ∼νbound − νfree) and intermediate (koff ∼ νbound − νfree) chemical exchange limits, where the second case is frequently encountered in 19F detection due to the large chemical shift dispersion of this nucleus. The fraction of the bound ligand (pb) for the [L]0 ≫ [P]0 case can be expressed by Fielding as42

pb =

[P]o Kd + [L]o

(3)

If one defines the differential frequency shift, γ, as

γ=

v1,obs − vfree v2,obs − vfree

(4)

Then the Kd value can be obtained using eqs 2 and 3 for two different ligand concentrations ([L]1 and [L]2) (see Supporting 685

dx.doi.org/10.1021/jm201441k | J. Med. Chem. 2012, 55, 678−687

Journal of Medicinal Chemistry

Article

of the carboxymethyl dextran surface of the CM5 chip treated in an identical manner as the other flow cell spot excluding the addition of protein. For Kd measurements the buffer was replaced with 50 mM sodium acetate pH 5.0, 150 mM NaCl, 0.005 (v/v) Tween20, and 5% (v/v) DMSO. Compound stocks prepared in DMSO (typically 20 mM) were serially diluted in running buffer and injected over the immobilized BACE-1. Association and dissociation time were typically set to 45 s and 90 s, respectively. The runs were performed at 25 °C with a flow rate from 10 to 90 μL/min and data collection rate of 10 Hz. The data were processed and analyzed using Scrubber-2 analysis software (BioLogic Software, Campbell, Australia). The sample response observed on the reference spot was subtracted from the sample response with immobilized BACE-1 to correct for systematic noise and baseline drift. Data was solvent corrected and the response from blank injections was used to double-reference the binding data. The data were molecular weight normalized and Kd values established using simple steady analysis with one global Rmax (or individual Rmax in cases where complete saturation was achieved). Chemistry. All compounds were purified to ≥95% purity as determined by reverse phase HPLC. HPLC analysis was obtained on Agilent 1100, using the following methods: HPLC method (3.6 min LC-MS run): Zorbax analytical C18 column (50 mm × 3 mm, 3.5 μm, 40 °C); mobile phase, A = 0.1% TFA in water, B = 0.1% TFA in acetonitrile; gradient, 0.0−3.6 min, 5−95% B; flow rate = 1.5 mL/min; 254 nm; 0 min post time; 1.0 μL injection.

Information for derivation):

Kd =

γ[L]1 − [L]2 1−γ

(5)

To obtain Kd values using this method, samples were usually prepared with L1 = 100 and L2 = 200 μM and with identical protein concentrations on the order of 1−5 μM. We refer to this method as dCSP. In the case of tighter binding ligands (koff ≪ νbound − νfree) (i.e., no chemical shift but line width changes of the observed resonance), the Kd values can be obtained in an analogous way from the differential line broadening of the observed free ligand signal at the two different concentrations. In this case, however, the changes of the line width are relatively much smaller and one needs to increase the protein concentration beyond the [L]0 ≫ [P]0 limit. Substituting the frequency shift in eq 4 for the half-width of the resonance line, one can write the following relationship:

γ=

p b1(1 − p b2 ) p b2 (1 − p b1)

where pb is the bound fraction of the ligand:

(6) 43

p b = {([L]o + [P]o + Kd) −

([L]o + [P]o + Kd)2 − 4[L]o [P]o }

/(2[L]o )



(7)

ASSOCIATED CONTENT

S Supporting Information *

The Kd value is then obtained from the numerical solution of eq 6. NMR Spectroscopy. All NMR experiments were performed on a Bruker Avance III NMR spectrometer (Bruker Biospin) operating at a 1 H frequency of 500.13 MHz using a SEF cryogenic probe equipped for direct 19F detection. One dimensional 19F spectra were acquired for each sample at 283 K using 1H decoupling with a spectra width of 71 428 Hz, an acquisition time of 917 ms, and 128 scans with a relaxation delay of 1 s. This yielded experiment times on the order of 4 min each, including 2 min for initial temperature equilibration. This experimental set up allowed all reference and screen spectra to be acquired in less than 24 h. Typically, a standard buffer (20 mM sodium phosphate, 50 mM NaCl (± DTT), pH = 7) is used in the screening campaigns. This appeared suitable for all our measured targets and allowed the elimination of buffer-dependent variations in the 19F chemical shifts while retaining good cryoprobe sensitivity at this moderate salt concentration. All data were processed using Topspin 2.1 (Bruker Biospin, Billerica, MA) and were then compared visually to the reference spectra using the spectral overlay feature. Hits were identified by signal intensity and/ or chemical shift changes. Since each compound in a pool had a unique chemical shift, hit identification was straightforward, and hit compounds could be identified by simply matching the chemical shift of the hit compound to that found in the compound database. Surface Plasmon Resonance Spectroscopy. Dissociation constant (Kd) measurements were performed on Biacore S51 and T100 instruments (GE Healthcare). BACE-1 protein and inhibitors for Biacore measurements were generated in-house; all other reagents were purchased from GE Healthcare or Sigma-Aldrich. Glycosylated BACE-1 was reacted with sodium periodate to oxidize cis-diols groups on sugar chains to aldehydes. The oxidized BACE-1 was immobilized at high density (10000−12000 RU) onto CM5 chips using aldehyde coupling chemistry and resulted in surface activities close to 100% based on reference inhibitor binding. The immobilization running buffer consisted of 10 mM HEPES pH 7.4 with 150 mM NaCl and immobilization steps consisted of a 3−4 min EDC/NHS activation step [200 mM 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride, 50 mM N-hydroxysuccinimide], 7 min 5 mM carbohydrizide in water, 7 min 1 M ethanolamine hydrochloride pH 8.5, 10 min 20 μg/mL oxidized BACE-1 in 10 mM sodium acetate pH 4.0, 7 min 100 mM sodium cyanoborohydride in 100 mM sodium acetate pH 4.0, 3 × 30 s 50 mM glycine pH 9.5 and 3 × 30 s 1 M NaCl in 100 mM NaHCO3 pH 9.5. The reference spot for all SPR experiments consisted

Derivations of eqs 3 and 5 as used in the manuscript. This material is available free of charge via the Internet at http:// pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*Tel: 805-313-5647. Fax: 805-498-9057. E-mail: jbjordan@ amgen.com (J.B.J.); [email protected] (L.P.).



ACKNOWLEDGMENTS The authors would like to acknowledge Hanh Nguyen for synthesis of some of the compounds used in this work. We would also like to acknowledge Dr. Richard Kriwacki (St. Jude Children’s Research Hospital) and Dr. Geoffrey Mueller (NIEHS) for providing protein with which to perform the druggability studies.



ABBREVIATIONS USED



REFERENCES

FBDD, fragment based drug discovery; FBS, fragment based screening; LE, ligand efficiency; HA, heavy atoms; HTS, high throughput screening; NMR, nuclear magnetic resonance; SPR, surface plasmon resonance; LEF, local environment of fluorine; BACE-1, β-site APP cleaving enzyme-1; ILNOE, interligand nuclear Overhauser enhancement; ACD, Available Chemicals Directory; ACDSC, Available Chemicals Directory Screening Compounds; dCSP, differential chemical shift perturbation; NOESY, nuclear Overhauser enhancement spectroscopy; STD, saturation transfer difference; WaterLOGSY, water-ligand observe gradient spectroscopy; SAR, structure−activity relationship

(1) Murray, C. W.; Rees, D. C. The rise of fragment-based drug discovery. Nat. Chem. 2009, 1, 187−192.

686

dx.doi.org/10.1021/jm201441k | J. Med. Chem. 2012, 55, 678−687

Journal of Medicinal Chemistry

Article

(2) Erlanson, D. A.; Hansen, S. K. Making drugs on proteins: sitedirected ligand discovery for fragment-based lead assembly. Curr. Opin. Chem. Biol. 2004, 8, 399−406. (3) Rees, D. C.; Congreve, M.; Murray, C. W.; Carr, R. Fragmentbased lead discovery. Nat. Rev. Drug Discovery 2004, 3, 660−672. (4) Carr, R.; Jhoti, H. Structure-based screening of low-affinity compounds. Drug Discovery Today 2002, 7, 522−527. (5) Nienaber, V. L.; Richardson, P. L.; Klighofer, V.; Bouska, J. J.; Giranda, V. L.; Greer, J. Discovering novel ligands for macromolecules using X-ray crystallographic screening. Nat. Biotechnol. 2000, 18, 1105−1108. (6) Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand efficiency: a useful metric for lead selection. Drug Discovery Today 2004, 9, 430−431. (7) Hajduk, P. J.; Greer, J. A decade of fragment-based drug design: strategic advances and lessons learned. Nat. Rev. Drug Discovery 2007, 6, 211−219. (8) Dalvit, C. Ligand- and substrate-based 19F NMR screening: Principles and applications to drug discovery. Prog. Nucl. Magn. Reson. Spectrosc. 2007, 51, 243−271. (9) Poppe, L.; Harvey, T. S.; Mohr, C.; Zondlo, J.; Tegley, C. M.; Nuanmanee, O.; Cheetham, J. Discovery of ligands for Nurr1 by combined use of NMR screening with different isotopic and spinlabeling strategies. J. Biomol. Screen. 2007, 12, 301−311. (10) Vulpetti, A.; Hommel, U.; Landrum, G.; Lewis, R.; Dalvit, C. Design and NMR-based screening of LEF, a library of chemical fragments with different local environment of fluorine. J. Am. Chem. Soc. 2009, 131, 12949−12959. (11) Morgenthaler, M.; Schweizer, E.; Hoffmann-Roder, A.; Benini, F.; Martin, R. E.; Jaeschke, G.; Wagner, B.; Fischer, H.; Bendels, S.; Zimmerli, D.; Schneider, J.; Diederich, F.; Kansy, M.; Muller, K. Predicting and tuning physicochemical properties in lead optimization: amine basicities. ChemMedChem 2007, 2, 1100−1115. (12) Leroux, F. Atropisomerism, biphenyls, and fluorine: a comparison of rotational barriers and twist angles. ChemBiochem 2004, 5, 644−649. (13) Gerebtzoff, G.; Li-Blatter, X.; Fischer, H.; Frentzel, A.; Seelig, A. Halogenation of drugs enhances membrane binding and permeation. ChemBiochem 2004, 5, 676−684. (14) Purser, S.; Moore, P. R.; Swallow, S.; Gouverneur, V. Fluorine in medicinal chemistry. Chem. Soc. Rev. 2008, 37, 320−330. (15) Hopkins, A. L.; Groom, C. R. The druggable genome. Nat. Rev. Drug Discovery 2002, 1, 727−730. (16) Barelier, S.; Krimm, I. Ligand specificity, privileged substructures and protein druggability from fragment-based screening. Curr. Opin. Chem. Biol. 2011, 15, 469−474. (17) Kubinyi, H. Drug research: myths, hype and reality. Nat. Rev. Drug Discovery 2003, 2, 665−668. (18) Brown, D.; Superti-Furga, G. Rediscovering the sweet spot in drug discovery. Drug Discovery Today 2003, 8, 1067−1077. (19) Miranker, A.; Karplus, M. Functionality maps of binding sites: a multiple copy simultaneous search method. Proteins 1991, 11, 29−34. (20) Laskowski, R. A. SURFNET: a program for visualizing molecular surfaces, cavities, and intermolecular interactions. J. Mol. Graph. 1995, 13 (323−30), 307−308. (21) Hendlich, M.; Rippmann, F.; Barnickel, G. LIGSITE: automatic and efficient detection of potential small molecule-binding sites in proteins. J. Mol. Graph. Model. 1997, 15, 359−363. (22) Liang, J.; Edelsbrunner, H.; Fu, P.; Sudhakar, P. V.; Subramaniam, S. Analytical shape computation of macromolecules: I. Molecular area and volume through alpha shape. Proteins 1998, 33, 1−17. (23) Liang, J.; Edelsbrunner, H.; Fu, P.; Sudhakar, P. V.; Subramaniam, S. Analytical shape computation of macromolecules: II. Inaccessible cavities in proteins. Proteins 1998, 33, 18−29. (24) Perot, S.; Sperandio, O.; Miteva, M. A.; Camproux, A. C.; Villoutreix, B. O. Druggable pockets and binding site centric chemical space: a paradigm shift in drug discovery. Drug Discovery Today 2010, 15, 656−667.

(25) Chen, I. J.; Hubbard, R. E. Lessons for fragment library design: analysis of output from multiple screening campaigns. J. Comput.-Aided Mol. Des. 2009, 23 (8), 603−620. (26) Cheng, A. C.; Coleman, R. G.; Smyth, K. T.; Cao, Q.; Soulard, P.; Caffrey, D. R.; Salzberg, A. C.; Huang, E. S. Structure-based maximal affinity model predicts small-molecule druggability. Nat. Biotechnol. 2007, 25, 71−75. (27) Hajduk, P. J.; Huth, J. R.; Fesik, S. W. Druggability indices for protein targets derived from NMR-based screening data. J. Med. Chem. 2005, 48, 2518−2525. (28) Edfeldt, F. N.; Folmer, R. H.; Breeze, A. L. Fragment screening to predict druggability (ligandability) and lead discovery success. Drug Discovery Today 2011, 16, 284−287. (29) Davies, T. G., Monfort, R. L. M., Williams, G., Jhoti, H. Pyramid: An integrated platform for fragment-based drug discovery. In Fragment-based Approaches in Drug Discovery; Erlanson, D. A., Jahnke, W., Ed.; Oxford University Press: Oxford, UK, 2006; pp 193−214. (30) London, R. E. Theoretical analysis of the inter-ligand overhauser effect: a new approach for mapping structural relationships of macromolecular ligands. J. Magn. Reson. 1999, 141, 301−311. (31) Murray, C. W.; Callaghan, O.; Chessari, G.; Cleasby, A.; Congreve, M.; Frederickson, M.; Hartshorn, M. J.; McMenamin, R.; Patel, S.; Wallis, N. Application of fragment screening by X-ray crystallography to beta-secretase. J. Med. Chem. 2007, 50, 1116−1123. (32) Bondi, A. Van der Waals volumes and radii. J. Phys. Chem. 1964, 68, 441−451. (33) Pauling, L. The Nature of the Chemical Bond, 6th ed.; Cornell University Press: Ithaca, NY, 1948. (34) Cheng, Y.; Judd, T.; Bartberger, M.; Chen, K.; Fremeau, R.; Hickman, D.; Hitchcock, S.; Jordan, J.; Li, V.; Lopez, P.; Louie, S.; Luo, Y.; Michelsen, K.; Nixey, T.; Powers, T.; Rattan, C.; Sickmier, E. St.; Jean, D.; Wahl, R.; Wen, P.; Wood, S. From fragment screening to in vivo efficacy: optimization of a series of 2-aminoquinolines as potent inhibitors of BACE1. J. Med. Chem. 2011, 54 (16), 5836−5857. (35) Massefski, W.; Redfield, A. Elimination of multiple-step spin diffusion in two-dimensional NOE spectroscopy of nucleic acids. J. Magn. Reson. 1988, 78, 150−155. (36) Dalvit, C.; Gossert, A. D.; Coutant, J.; Piotto, M. Rapid acquisition of 1H and 19F NMR experiments for direct and competition ligand-based screening. Magn. Reson. Chem. 2011, 49, 199−202. (37) Geschwindner, S.; Olsson, L. L.; Albert, J. S.; Deinum, J.; Edwards, P. D.; de Beer, T.; Folmer, R. H. Discovery of a novel warhead against beta-secretase through fragment-based lead generation. J. Med. Chem. 2007, 50, 5903−5911. (38) Halgren, T. A. Identifying and characterizing binding sites and assessing druggability. J. Chem. Inf. Model. 2009, 49, 377−389. (39) Lipkus, A. A proof of the triangle inequality for the Tanimoto distance. J. Math. Chem. 1999, 26, 263−265. (40) Durant, J. L.; Leland, B. A.; Henry, D. R.; Nourse, J. G. Reoptimization of MDL keys for use in drug discovery. J. Chem. Inf. Comput. Sci. 2002, 42, 1273−1280. (41) Swift, T. J.; Connick, R. E. NMR-relaxation mechanisms of O17 in aqueous solutions of paramagnetic cations and the lifetime of water molecules in the first coordination sphere. J. Chem. Phys. 1962, 37, 307−320. (42) Fielding, L. NMR methods for the determination of proteinligand dissociation constants. Prog. Nucl. Magn. Reson. Spectrosc. 2007, 51, 219−242. (43) Forsen, S.; Hoffman, R. Study of moderately rapid chemical exchange reactions by means of nuclear magnetic double resonance. J. Chem. Phys. 1963, 39, 2892.

687

dx.doi.org/10.1021/jm201441k | J. Med. Chem. 2012, 55, 678−687